The new AI economy may reverse the promise of automation. Instead of AI working for you, you may end up working for AI agents that assign tasks, judge your performance, and decide your bonuses. This article lays out the evidence, the historical precedents, and the two realistic ways to stay on the controlling side of that relationship.
What Is the New AI Economy?
The new AI economy is a labor market where AI agents, software systems that act and decide on their own, direct human work instead of only assisting it. The thesis comes from Bruno Faggion's March 2026 video on the subject: people expected AI to work for them, but the first jobs agents take may be management jobs, leaving humans to execute tasks for machines.
The video sketches two classes of people. One group builds, maintains, and sets the directives for AI systems, either as developers or as company owners. The other group performs the tasks agents delegate and answers to those agents. In this model, the middle of the pyramid, the traditional managerial layer, is the part that disappears first.
The speaker frames it with a joke: not even Karl Marx, in his wildest dreams and nightmares, imagined a scenario this strange, where robots are bosses and most humans answer to them. This is an argument, not a measured trend. The video supports it with two early examples and one historical precedent, which the sections below examine in order.
Hire a Human: When AI Agents Become Employers
Around early 2026, a developer launched a project called Hire a Human, a site where AI agents pay humans to do things machines still cannot. The examples given in the video: pick up a package from a specific store, or stand outside a shop for an hour counting how many people walk in. The agents handle the reasoning and the payment; the human provides physical presence.
The project is early-stage and small, so it proves possibility, not scale. Its significance is directional. Gig platforms already route work through algorithms, and Hire a Human removes the last human interface: the requester itself becomes software. If you have made videos about shifts like this, tools such as Skala Blog can turn that footage into searchable written articles.
Automation 1.0: The Algorithm Is Already Your Boss
The strongest precedent for the new AI economy already exists. Uber drivers and iFood couriers take orders from matching algorithms. A central system assigns rides and deliveries, and the worker has little control over which fare or which restaurant comes next. Drivers feel like their own bosses because they choose their hours, but the algorithm chooses the work.
The video calls this era Automation 1.0: the algorithm performs exactly one function, matching supply to demand. It cannot ask a driver to help a passenger with accounting mid-ride, or to do a backflip on arrival. Because the system is narrow, its power over workers stays narrow too. The speaker deliberately takes no side on whether this model is good or bad for workers; the point is structural.
Automation 2.0 removes that limit. An agent can review your work, score it, compare it to peers, and recommend consequences. The video cites a support worker at a Brazilian food-delivery startup, one of the companies clustered in the so-called Brazilian Silicon Valley around Faria Lima, whose manager was reportedly replaced by an agent: to issue a customer refund, she now has to convince the AI that the refund is justified. This is a single anecdote from Twitter, so treat it as illustration, but it shows the shape of the change.
| Dimension | Automation 1.0 | Automation 2.0 |
|---|---|---|
| What the system does | Matches supply to demand, one function | Reviews, scores, and directs work end to end |
| Role over the worker | Dispatches tasks | Acts as a manager and judge |
| Example | Uber and iFood ride matching | AI agent deciding whether a refund is justified |
| Human oversight | Humans run the company above it | Agents may sit in the managerial layer itself |
Why Management Is the First Layer AI Can Replace
A human manager passes down directives from directors, evaluates performance, sets daily plans and weekly targets, and trains new staff. The video's argument is that each of these tasks maps cleanly onto what agentic systems already attempt: analyzing spreadsheets, reading conversations, scoring output, and proposing decisions.
Tools mentioned in the video point that direction. Anthropic Claude introduced agentic collaboration features such as Claude Cowork, and workflow builders like n8n, a source-available automation platform with self-hosting options, let people assemble multi-step agents without heavy engineering. The video also references Claude agentic bots as early examples of systems that can act on their own.
The pace argument comes from ChatGPT. ChatGPT, OpenAI's assistant, arrived in November 2022. At launch it was impressive but unusable for real work. Within roughly four years it became a working tool for programmers and lawyers, and reportedly for some doctors and judges who use it to avoid reading hundreds of pages of case files, with one figure cited being processes of up to 500 pages. The video's inference: agents stand today where ChatGPT stood in 2022, and could take on managerial functions within two or three years.
The structural consequence worries the speaker most. In the economy of the 1960s through 1980s, a person could enter at the bottom, become a coordinator, then a supervisor, then a manager, and occasionally a director. If agents hold the managerial layer, that ladder develops a gap. Ambitious workers can no longer climb by doing the jobs AI does for free. He concedes the ladder was already weakening before agents arrived, but the 70s-style corporate ascent was at least possible. For support roles in a SaaS company, an agent could read every conversation, give feedback, judge bonuses, and flag below-average performers to human executives, all tasks he considers straightforward for current technology.
Which Jobs Resist AI Automation
The video does not claim every job falls. It names categories likely to automate first and categories likely to need humans for years.
Likely to automate early:
- WhatsApp and chat-based customer support, where agents already handle conversations.
- Call centers, which subject-matter specialists expect to automate soon.
- Sales of inexpensive products, where human rapport matters less than in high-ticket sales.
Likely to need humans longer:
- Retail store staff, because shoppers want a person, not a speaker playing an AI voice.
- Physical work such as assembling a McDonald's hamburger, at least until robotics advances considerably.
- Many office roles where presence and judgment still matter.
The pattern is task-dependent, not industry-dependent. Jobs combining physical presence with human trust resist longest; anything that lives in text and follows reviewable rules is exposed.
Two Paths to Stay on Top in the New AI Economy
The video closes with two ways to land in the controlling class rather than the managed one.
- Build or operate AI agents. If you have a technical orientation, learning to create and manage automations puts you on the side that issues directives. Platforms such as n8n lower the entry barrier for this path.
- Build a lean business. The speaker argues entrepreneurship has gotten cheaper in two waves: the internet removed the cost of physical premises and inventory, and AI agents now cut the other major cost, payroll. Robots work with you rather than for someone else.
For readers pursuing the technical path, structured programs exist outside the video too, such as the Bootcamp do Dev Doido by Gustavo Dev Doido and the Crazystack Typescript course at crazystack.com.br, which teach the kind of development and automation skills this path assumes.
Both paths rest on inference from early evidence, not on measured outcomes. But they share a logic: in a market split between people who direct agents and people who answer to them, ownership of the agents is what decides your side of the split.
Will AI agents really become managers?
It is plausible but unproven. Managerial tasks such as reviewing conversations, scoring performance, and setting targets match what current agentic tools attempt, so the trajectory is credible. Whether companies actually trust agents with hiring, bonuses, and firing remains an open governance question.
What is the Hire a Human project?
A small developer project from around 2026 where AI agents pay humans for physical errands machines cannot perform, such as counting customers at a store. It demonstrates the direction of the new AI economy, not its scale.
Is Uber an example of working for an algorithm?
Partially. Uber drivers choose when to work, but the Uber matching algorithm decides which rides they get. The video treats this narrow dispatch automation as a precedent for agents with far broader managerial powers.
What is the difference between Automation 1.0 and Automation 2.0?
Automation 1.0 describes pre-AI algorithms that do one job, such as matching a driver to a passenger. Automation 2.0 describes AI-era agents that can evaluate work, give feedback, and recommend consequences. The first allocates labor; the second supervises it.
Which jobs will AI automate first?
The video names chat-based support, call centers, and sales of cheap products as the earliest to fall. All three live in text, follow rules, and do not require physical presence or personal trust.
Which jobs are safest from AI agents?
Retail staff, physical food preparation, and many office roles that depend on presence and judgment. A shopper at a mall wants to talk to a person, and a hamburger still needs human hands until robotics catches up.
Should I learn to build AI automations?
If you want to be in the group that controls agents rather than answers to them, it is one of two paths the video recommends, alongside starting a lean business that uses agents to cut payroll costs. Both are strategic bets, not guarantees.
Why does the managerial layer matter so much in this argument?
Because it is the rung ordinary workers climb. If agents occupy coordinator, supervisor, and manager roles, the corporate ladder loses its middle, and entry-level workers lose the route to director roles that existed from the 1960s onward.
Turn Your Own Video Arguments Into Written Articles
The argument above started as a 15-minute YouTube video, and its thesis about who controls the agents of the new AI economy deserves to reach readers who never watch video. If you have the same kind of thinking stored in your own recordings, Skala Blog converts it: paste a YouTube URL, get the transcript, and generate a structured written article you can review and publish. Your video knowledge becomes searchable text without you rewriting a word.
Fork this article
Start a new branch from the same video, shaped your way. You keep the credit; the original keeps the attribution.
A fork in another language is filed as a translation of this article, so the two pages point at each other. You can unlink it later from the editor.
0/240
You are creating
- Format
- For
- Language
- Source
- Your angle
You will be asked to sign in before it is generated.
Buy credits